We have developed methods for the determination of cell-specific gene expression for the four major cell types that compose prostate cancer including stroma cells. These studies reveal that hundreds of significant differential expressions distinguish normal stroma from tumor-adjacent stroma. These differences have been exploited to develop a Diagnostic Classifier based on genes expressed nearly exclusively in stroma. The Diagnostic Classifier accurately identifies the presence or absence of tumor in over 300 independent prostate cancer test sets including subsets containing on1 of microarray data based one. Further a subset of the Classifier genes accurately identifies "reactive" stroma vs. normal stroma that is associated with poor outcome indicating prognostic potential. We hypothesize that this stroma-based Classifier may be applied to negative clinical biopsies of patients 1 slatted for repeat biopsy to determine the presence of tumor thereby providing Early Detection 6 - 12 months ahead of the repeat biopsy results. Since most of the annual one million biopsies are negative and ~190,000 are scheduled for repeat biopsy, this is an enormous unmet need with no accepted biomarkers for diagnosis and/or early detection for negative biopsy cases. In Aims 1 and 2 validations with independent test array data and validation with LCM-prepared samples are proposed. Aim 3 is a prospective clinical study of new patients consented via the UCI SPECS consortium and SPORE at Northwestern University to collect negative biopsies of patients slatted for repeat biopsy. The repeat biopsy results are used to test the hypothesis that 72 predicted probe sets are valid Diagnostic, Early Detection, and Prognostic biomarkers. A clinical prospective study of Early Detection is possible since the analysis for each consented patient is completed at the time of the repeat biopsy. This is a novel and rare use of negative biopsy material. Therefore all analyses of Aim 3 will be done by genome wide expression analysis to determine the expression of the hypothetical genes and provide for identification of other probe sets of genes that meet the training and testing criteria for new members of the Classifier. The last Aim, 4, is to test the hypothesis that tumor-derived TGFBetal expression is a factor that is associated with promoting differential gene expression of many of the prognostic genes observed here. Thus a program to validate, extend, and understand an important aspect of mechanism of a profile of genes that are diagnostic based on stroma expression alone with subsets of genes with Early Detection and Prognostic potential is proposed. Key interactions with the EDRN for sample supplementation and as a site of a blinded validation are proposed. The anticipated test development is planned and all key data generated here will be developed in a CLIA lab.